Note: Descriptions are shown in the official language in which they were submitted.
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
EVALUATING THE SAFETY PERFORMANCE OF
VEHICLES
BACKGROUND
This application claims priority to and the benefit of U.S. Patent Application
No. 16/521,856, filed July 25, 2019, the entire contents of which is
incorporated here
by reference.
BACKGROUND
This description relates to evaluating the safety performance of vehicles.
Autonomous vehicles hold the promise of being safer than conventional
vehicles driven by humans. To assess the relative safety performance of
autonomous
vehicles and conventional vehicles, industry professionals and regulatory
bodies use
.. safety-correlated metrics such as crash rate, which measures the number of
vehicle
crashes in which a vehicle has been involved relative to the number of miles
traveled
by the vehicle.
Because conventional vehicles have recorded many more miles than
autonomous vehicles, and thus have been exposed to significantly higher risks
of
vehicle crashes, autonomous vehicle producers are racing to accumulate miles
to
obtain statistically meaningful, exposure-based measurements of vehicle safety
performance for comparison with conventional vehicles.
SUMMARY
In general, in an aspect, safety performance of vehicles is monitored or
evaluated, or both. Sensor data is received from one or more sensors at a
vehicle. One
or more risk sources encountered by the vehicle during a period of operation
of the
vehicle are identified based on the sensor data. One or more safety events
experienced
by the vehicle during the period of operation are identified based on the
sensor data. A
risk of a safety event experienced by the vehicle during the period of
operation is
determined based on the one or more risk sources. A safety performance of the
vehicle
during the period of operation is determined based on the one or more safety
events
and the risk.
1
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
The risk can be measured by a risk score. The period of operation of the
vehicle can be associated with a distance traveled by the vehicle. The period
of
operation of the vehicle can be associated with a time of operation of the
vehicle. The
sensors can be included in one or more telematics devices at the vehicle.
The one or more risk sources can include environmental risk sources due to
conditions or features of the environment in which the vehicle is operated.
The
environmental risk sources can include other vehicles in the environment. The
one or
more safety events can include a crash of the vehicle or near-crash of the
vehicle, or
both. Determining the risk can include determining a risk of a safety event
posed by
each of the risk sources.
Risk sources due to the conduct of a driver of the vehicle can be filtered
from
the one or more risk sources. A sum of the risk of the safety event posed by
each of
the risk sources can be calculated to determine the risk. The sum of the risk
can be a
weighted sum. Determining the safety performance of the vehicle can include
determining a sum of the one or more safety events relative to the risk. The
sum of the
one or more safety events relative to the risk can be a weighted sum. Each of
the one
or more safety events can be weighted based on the severity of the safety
event
The risk can be applied to a distance traveled by the vehicle during the
period
of operation of the vehicle to produce a risk-normalized distance traveled by
the
.. vehicle. The safety performance of the vehicle during the period of
operation of the
vehicle can be determined based on the one or more safety events and the risk-
normalized distance traveled,. The risk can be normalized based on a standard
level of
risk experienced for the period of operation of the vehicle. A safety
performance of a
second vehicle in a same environment as the vehicle during at least part of
the period
of operation of the vehicle can be determined based on the sensor data. The
safety
performance of the second vehicle can be determined based on identifying one
or
more risk sources experienced by the second vehicle in the same environment as
the
vehicle during at least part of the period of operation of the vehicle. One or
more
safety events experienced by the second vehicle during at least part of the
period of
operation of the second vehicle can be identified based on the sensor data,.
A risk of a safety event experienced by the second vehicle during at least
part
of the period of operation of the vehicle can be determined based on the one
or more
2
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
risks experienced by the second vehicle. A relative safety performance of the
vehicle
during at least part of the period of operation of the vehicle can be
determined by
comparing the safety performance of the vehicle with the safety performance of
the
second vehicle. The risk for the vehicle can be combined with a risk of a
safety event
for one or more other vehicles having a same vehicle type as the vehicle. A
safety
performance of the vehicle type can be determined based in part on the
combined risk.
In general, in an aspect, a system including one or more processors is
provided. The system includes computer storage storing executable computer
instructions executable by the one or more processors to receive sensor data
from one
or more sensors at a vehicle. The instructions are executable by the one or
more
processors to identify, based on the sensor data, one or more risk sources
encountered
by the vehicle during a period of operation of the vehicle. The instructions
are
executable by the one or more processors to identify, based on the sensor
data, one or
more safety events experienced by the vehicle during the period of operation
of the
vehicle. The instructions are executable by the one or more processors to
determine,
based on the one or more risk sources, a risk of a safety event experienced by
the
vehicle during the period of operation of the vehicle. The instructions are
executable
by the one or more processors to determine, based on the one or more safety
events
and the risk, a safety performance of the vehicle during the period of
operation of the
vehicle.
Determining the safety performance of the vehicle can include determining a
sum of the one or more safety events relative to the risk. The computer
storage can
store instructions executable by the one or more processors to apply the risk
to a
distance traveled by the vehicle during the period of operation of the vehicle
to
produce a risk-normalized distance traveled by the vehicle, and determine,
based on
the one or more safety events and the risk-normalized distance traveled, the
safety
performance of the vehicle during the period of operation of the vehicle.
The computer storage can store instructions executable by the one or more
processors to identify, based on the sensor data, one or more risks
experienced by a
second vehicle in a same environment as the vehicle during at least part of
the period
of operation of the vehicle. The computer storage can store instructions
executable by
the one or more processors to identify, based on the sensor data, one or more
safety
3
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
events experienced by the second vehicle during at least part of the period of
operation of the vehicle. The computer storage can store instructions
executable by
the one or more processors to determine, based on the one or more risks
experienced
by the second vehicle, a risk of a safety event experienced by the second
vehicle
during at least part of the period of operation of the vehicle. The computer
storage can
store instructions executable by the one or more processors to determine,
based on the
one or more safety events experienced by the second vehicle and the risk for
the
second vehicle, a safety performance of the second vehicle during at least
part of the
period of operation of the vehicle.
In general, in an aspect, a computer-readable storage medium having
instructions executable by one or more processors is provided. The
instructions are
executable by the one or more processors to cause the processors to perform
one or
more operations. The one or more operations include receiving sensor data from
one
or more sensors at a vehicle. The one or more operations include identifying,
based on
the sensor data, one or more risk sources encountered by the vehicle during a
period
of operation of the vehicle. The one or more operations include identifying,
based on
the sensor data, one or more safety events experienced by the vehicle during
the
period of operation of the vehicle. The one or more operations include
determining,
based on the one or more risk sources, a risk of a safety event experienced by
the
vehicle during the period of operation of the vehicle. The one or more
operations
include determining, based on the one or more safety events and the risk, a
safety
performance of the vehicle during the period of operation of the vehicle.
The one or more operations can include applying the risk to a distance
traveled
by the vehicle during the period of operation of the vehicle to produce a risk-
normalized distance traveled by the vehicle. The one or more operations can
include
determining, based on the one or more safety events and the risk-normalized
distance
traveled, the safety performance of the vehicle during the period of operation
of the
vehicle. The one or more operations can include identifying, based on the
sensor data,
one or more risks experienced by a second vehicle in a same environment as the
vehicle during at least part of the period of operation of the vehicle. The
one or more
operations can include identifying, based on the sensor data, one or more
safety
events experienced by the second vehicle during at least part of the period of
4
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
operation of the vehicle. The one or more operations can include determining,
based
on the one or more risks experienced by the second vehicle, a risk of a safety
event
experienced by the second vehicle during at least part of the period of
operation of the
vehicle. The one or more operations can include determining, based on the one
or
more safety events experienced by the second vehicle and the risk for the
second
vehicle, a safety performance of the second vehicle during at least part of
the period
of operation of the vehicle.
In general, in an aspect, a system including one or more processors and a
plurality of modules is provided. The system includes a risk identification
module
executable by the one or more processors to identify, based on sensor data
from one or
more sensors at a vehicle, one or more risk sources encountered by the vehicle
during
a period of operation of the vehicle. The system includes a safety event
identification
module executable by the one or more processors to identify, based on the
sensor data,
one or more safety events experienced by the vehicle during the period of
operation of
the vehicle. The system includes a risk scoring module executable by the one
or more
processors to determine, based on the one or more risk sources, a risk of a
safety event
experienced by the vehicle during the period of operation of the vehicle. The
system
includes a safety performance evaluation module executable by the one or more
processors to determine, based on the one or more safety events and the risk,
a safety
performance of the vehicle during the period of operation of the vehicle.
In general, in an aspect, a system is provided that includes a device in a
vehicle configured to capture operating period information indicating the
extent or
duration of one or more periods of operation of the vehicle. The device is
configured
to capture safety information indicative of risks and safety events involving
the
vehicle or a vehicle in the vicinity of the vehicle during each of the one or
more
periods of operation of the vehicle. The device is configured to send the
operating
period information and the safety information to a server for processing to
determine
a safety performance of the vehicle based on a risk score and the safety
events.
These and other aspects, features, and implementations can be expressed as
methods, apparatus, systems, components, program products, methods of doing
business, means or steps for performing a function, and in other ways, and
will
become apparent from the following descriptions, including the claims.
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
DESCRIPTION
FIG. 1 is a schematic diagram illustrating a vehicle safety performance
evaluation technology 100.
FIG. 2 is a block diagram depicting an example server 130 for evaluating
vehicle safety performance.
FIG. 3 is a flowchart depicting a method for monitoring and evaluating the
safety performance of vehicles.
Each time a vehicle operates, it may be exposed to certain situations or
conditions that put the safety of the vehicle, its users, or the surrounding
public at risk.
For example, a vehicle may encounter risk due to features or conditions of the
environment in which it operates, such as the weather or the proximity of
other
vehicles, risk due to conduct by the driver of the vehicle (for example,
speeding,
distracted driving, and so forth), risk due to the vehicle itself (for
example, damage to
or malfunctioning of the vehicle, and so forth. We sometimes refer to the
situations or
conditions that produce risk as "risk sources," and we use the term "risk"
broadly to
refer to the likelihood of injury, loss, or another undesirable outcome due to
one or
more risk sources.
Fortunately, the majority of risk that a vehicle encounters may be avoided or
otherwise managed by the vehicle or its driver. In some cases, however, risk
can
materialize in the form of a vehicle crash, a near-crash, or another unsafe
condition at
the vehicle, which we sometimes refer to collectively as a "safety event."
To measure the safety of a vehicle or its driver, or both (sometimes referred
to
as the "safety performance" of the vehicle or its driver), it may be useful to
understand how the vehicle or its driver responds when exposed to the risk of
a safety
event. For example, vehicle crash rate seeks to measure the safety performance
of a
vehicle by calculating the number of reported vehicle crashes relative to the
vehicle-
miles traveled (VMT). In this way, vehicle crashes can serve as a measure of
the
materialization of risk at the vehicle, and VMT can serve as a proxy for the
risk
experienced by the vehicle.
However, vehicle crash rate can be an imprecise measure of safety
performance. Many vehicle crashes may go unreported, and measuring safety
based
solely on vehicle crashes can fail to consider other safety events, such as
near-crashes,
6
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
that can be important for measuring the safety performance of the vehicle or
the
driver, or both. Further, the risk of a safety event experienced by the
vehicle typically
varies from mile to mile. For example, operating a vehicle on a rural road at
midnight
under normal weather conditions may pose substantially less risk per mile than
operating the same vehicle on a congested urban road during rush hour when it
is
snowing. As a result, using VMT as a proxy for the risk experienced by a
vehicle,
without more, may produce inaccurate or biased measurements of the vehicle's
safety
performance.
The vehicle safety performance evaluation technology described in this
specification (sometimes referred to as "the technology") can acquire and
process
sensor data generated at a vehicle to evaluate the safety performance of the
vehicle or
other vehicles in the surrounding environment. To do so, the technology can
receive
sensor data from one or more sensors installed in the vehicle, included in a
telematics
device (for example, a smartphone) brought into the vehicle, or both. The
technology
can process the acquired sensor data to identify the sources of risk
encountered by the
vehicle and the safety events experienced by the vehicle during operation. In
some
cases, the technology is capable of using the identified risk sources, safety
events for
the vehicle, and safety events of other vehicles to establish the risk of a
safety event
posed by each risk source. Based on the identified risk sources and the
established
risk, the technology can generate a risk score that corresponds to the risk of
a safety
event experienced by the vehicle during a particular period of operation, such
as a
distance traveled or time operated. Because the generated risk score can
account for
both the actual sources of risk encountered by the vehicle and the risk of a
safety
event posed by each source, the technology, when compared with conventional
systems, can generate a risk score that serves as a more accurate measure of
the risk
experienced by the vehicle during the period of operation.
The technology can use the risk score to evaluate the safety performance of a
vehicle. In some cases, the technology can determine the safety performance of
a
vehicle by calculating the number of safety events experienced by the vehicle
over a
particular period of operation relative to the risk score for the vehicle over
the same
period. For example, the technology may determine from the sensor data that,
over a
distance of one mile, the vehicle encountered ten risk sources and experienced
one
7
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
safety event in the form of a near-crash. Based on the ten identified risk
sources and
the risk of a safety event posed by each source, the technology can determine
the risk
of a safety event for the vehicle over the mile, which can be represented as a
risk
score. Assuming, for example, that the risk score for the vehicle over the
mile is
determined to be 0.05 (indicating a 5% chance of a safety event), the
technology can
divide the number of safety events (1) by the risk score (0.05) to determine
that the
safety performance of the vehicle over the mile is equal to 20. In this way,
safety
events can serve as a measure of the materialization of risk at the vehicle,
and the risk
score can serve as a measure of the risk of a safety event experienced by the
vehicle.
By more accurately defining the vehicle's exposure to risk through the risk
score, the
technology can better approximate the safety performance of the vehicle
compared to
techniques that rely solely on VMT as a proxy for risk.
In some cases, the technology can evaluate the safety performance of the
vehicle or vehicle type over multiple periods of operation. We use the term
"vehicle
type" broadly to include any set of two or more vehicles that share a common
characteristic, such as a vehicle make, a vehicle model, an autonomous
vehicle, a
conventional vehicle, or a vehicle class, such as a sedan, a SUV, or a truck,
among
others. To do so, the technology can combine multiple risk scores for a
vehicle or
vehicle type to determine a combined risk score that represents the total risk
of a
safety event encountered by the vehicle or vehicle type over multiple periods
of
operation. The technology can then calculate the combined number of safety
events
experienced by the vehicle or vehicle type relative to the combined risk score
to
determine the safety performance of the vehicle or vehicle type. By combining
the
safety events and risk scores, statistically significant metrics of safety
performance
can be determined and compared among different vehicles or vehicle types. For
example, when applied to all travel by conventional vehicles, the technology
can
evaluate the safety performance of all conventional vehicles and their human
drivers.
When applied to all travel by autonomous vehicles, the technology can evaluate
the
safety performance of all autonomous vehicles and their autonomous (for
example,
computer-implemented) drivers. In this way, the technology can assess the
relative
safety performance of conventional vehicles, autonomous vehicles, and their
human
or computer-implemented drivers.
8
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
In some cases, it may be useful to compare the safety performance metrics
described here with conventional metrics, such as those that use VMT as a
proxy for
risk. Accordingly, the technology can normalize the risk score to a standard
or average
risk of a safety event for a particular period of operation, such as a mile
driven under a
fixed level or risk. The technology can apply the normalized risk score to
produce, for
example, a normalized distance traveled by the vehicle or vehicle type based
on the
actual risk encountered by the vehicle or vehicle type. In this way, low-risk
travel,
such as driving one mile in the desert at 2 a.m., may result in a reduced risk-
normalized distance traveled (for example, 0.1 standard miles), whereas high-
risk
mile travel, such as driving one mile in Manhattan, N.Y. during rush hour, may
result
in an increased risk-normalized distance traveled (for example, 3.5 standard
miles).
The technology can then calculate the number of safety events experienced by
the
vehicle or vehicle type relative to the risk-normalized distance traveled to
determine
the safety performance of the vehicle or vehicle type. This is analogous to
the
conventional metric in which vehicle crashes are taken as the numerator and
VMT is
taken as the denominator.
To normalize the distance traveled, the distances traveled in an environment
can be compared to a "typical" risk-normalized distance. In some
implementations, a
typical risk-normalized distance traveled refers to a "yearly average" mile.
In such
instances, risks are aggregated across several driving environments (which can
have
different attributes, similar attributes, or both) to determine a typical risk
for a given
location during a given time frame. For example, assume that one "yearly
average"
mile reflects the aggregated risk corresponding to 100 randomly (or pseudo-
randomly) selected one standard mile segments across the United States in the
year
2019. A one standard mile segment in the desert at 2 a.m. can be compared to
one
"yearly average" mile and can be reduced to a risk-normalized distance of 0.1
standard miles, reflecting that the risk of driving one standard mile in the
desert at 2
a.m. can be less risky than driving one average standard mile.
In some implementations, a typical risk-normalized distance traveled refers to
a "Cambridge" mile. In such instances, a canonical road segment and point in
time is
arbitrarily determined and the determined segment and point in time is defined
as a
"Cambridge" mile. For example, assume that one "Cambridge" mile reflects the
risk
9
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
of driving a selected one standard mile segment of Massachusetts Avenue in the
Porter Square section of Cambridge, MA on April 13, 2019 from 1:00 p.m. to
2:00
p.m. A one standard mile segment of Manhattan, N.Y. between 5:00 p.m. and 6:00
p.m. can be compared with one "Cambridge" mile, and can be increased to a risk-
normalized distance of 3.5 standard miles, reflecting that the risk of driving
one
standard mile in Manhattan between 5:00 p.m. and 6:00 p.m. can be more risky
that
driving one "Cambridge" mile.
In some implementations, the technology can use sensor data produced at a
vehicle to evaluate the safety performance of other vehicles in the
surrounding
environment. For example, the technology can process the sensor data produced
at the
vehicle to identify risk sources, safety events, and a risk score for the
vehicle and each
of the other vehicles in the surrounding environment. The technology can then
use the
safety events and the risk scores to evaluate the safety performance of the
vehicle and
each of the other vehicles in the surrounding environment. By evaluating the
safety
performance of both the vehicle and the other vehicles in the surrounding
environment, the technology can increase the amount of data collected and can
make
comparisons between the safety performance of the vehicle and the other
surrounding
vehicles. Such a comparison may be significant because the risk sources
experienced
by the vehicle and surrounding vehicles during operation (such as the weather,
the
road conditions, the amount of traffic, and so forth) can be similar.
FIG. 1 is a schematic diagram illustrating a vehicle safety performance
evaluation technology 100. Generally, the technology 100 can include a variety
of
components and devices that cooperate to perform the activities and functions
described in this specification. Each of the components and devices can
include one or
more hardware components, software applications, and data communication
channels
for communicating data between the hardware components and software
applications
of devices that are part of the technology. Some of the components and devices
can be
implemented as computer-readable storage mediums containing computer-readable
instructions for execution by one or more processors within the technology
100. The
technology 100 shown in FIG. 1 can include additional, fewer, or alternate
components, including those discussed previously and later.
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
Referring to FIG. 1, the technology 100 is capable of monitoring and
evaluating the safety performance of one or more vehicles 102 as they operate
within
an environment 104. Each of the vehicles 102 are capable of being operated by
a
driver 106 (e.g., a conventional vehicle), by an on-board computer 108 (e.g.,
an
autonomous vehicle), or by a combination of a driver 106 and an on-board
computer
108 (e.g., a semi-autonomous vehicle). However, the techniques described here
are
not limited to these vehicles, and we use the term "vehicle" broadly to
include, for
example, any kind of ground conveyance such as a car, truck, bus, bicycle,
motorcycle, or recreational vehicle, among others. The environment 104 may
include
the surroundings, such as the road, infrastructure, pedestrians, or other
vehicles, and
conditions, such as the time of day, the weather, or the level of lighting, in
which the
vehicle 102 operates.
As each vehicle 102 operates within the environment 104, it may encounter
one or more risk sources 110 that expose the vehicle, its users, or the
surrounding
public to risk.
For example, in some cases, the vehicle 102 may encounter one or more risk
sources 110 due to the environment 104 (sometimes referred to as
"environmental risk
sources"). The one or more risk sources 110 due to the environment can include
one
or more features or conditions of the environment such as, among others, the
weather,
the time of day, the level of lighting, the angle of lighting, the road type,
the road
condition, the area of the road in which the vehicle is operating (for
example, an
intersection, a merge lane, a shoulder, and so forth), the level of vehicle
traffic, the
level of pedestrian traffic, the speed of vehicles in the environment, the
geography, the
presence (or absence) of road markings, the presence (or absence) of
construction, the
proximity of the vehicle to other vehicles or objects, the speed of other
vehicles, or
the acceleration of other vehicles, the behavior of other vehicles.
In some cases, the vehicle 102 may be exposed to one or more risk sources
110 due to the driver 106 of the vehicle (sometimes referred to as "driver-
specific risk
sources"). For example, the one or more risk sources 110 due to the driver 106
can
include one or more of the following: distracted driving (for example, driving
while
talking on the phone), driving while incapacitated, speeding, hard
acceleration, hard
braking, hard cornering, drifting, or swerving.
11
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
In some cases, the vehicle 102 may be damaged or malfunctioning, and thus
may be a source of risk to itself or other vehicles in the environment 104
(sometimes
referred to as "vehicle-specific risk sources").
At times, the risk posed by one or more of the risk sources 110 encountered by
the vehicle 102 may materialize in the form of a safety event 111 (such as a
vehicle
crash, a near-crash, or another unsafe condition at the vehicle) that can
jeopardize the
safety of the users of the vehicle, other persons within the environment 104,
or both.
To monitor for the risk sources 110 and safety events 111 experienced by the
vehicle 102 or other vehicles within the environment 104, the vehicle 102 can
include
one or more sensors 112. In some implementations, the one or more sensors 112
include one or more of the following: one or more accelerometers, one or more
speed
sensors, or one or more position sensors (such as a Global Positioning System
(GPS)).
In such instances, the one or more sensors 112 can monitor one or more of the
following: acceleration, speed, or position of the vehicle 102. In some
implementations, the one or more sensors 112 include one or more of the
following:
one or more audio sensors (such as a microphone), or one or more image sensors
(such as an in-cabin camera). In such instances, the one or more sensors 112
can
monitor, for example, the driver 106 of the vehicle. In some implementations,
the one
or more sensors 112 include one or more of the following: one or more radar
systems,
one or more LIDAR systems, or one or more sonar systems. In such instances,
the one
or more sensors 112 can monitor, for example, the vehicle 102 or other
features in the
surrounding environment 104.
In general, the vehicle 102 can include any number or type of sensors 112
suitable to monitor the vehicle 102, its driver 106, and the surrounding
environment
104, including, among others, one or more of the following: accelerometers,
magnetometers, gyroscopes, inertial measurement units (IMUs), speed sensors,
position sensors (such as a GPS), barometric sensors, weight sensors, engine
sensors,
alternator sensors, odometer sensors, vibration sensors, voltage sensors,
oxygen
sensors, biometric sensors, light sensors, image sensors, audio sensors,
ultrasonic
sensors, electronic control unit (ECU) devices, radar systems, LIDAR systems,
or
sonar systems.
12
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
The sensors 112 may be installed in (or on) the vehicle 102, included in one
or
more telematics devices 114 brought into the vehicle, or both. Each of the one
or more
telematics devices 114 may include one or more processors 116 and memory 118
to
process and store the sensor data and other information, and a communications
interface 120 to enable wired or wireless communications with other components
or
devices of the technology, such as the sensors 112, the on-board computer 108,
and a
server 130. The telematics devices 114 can include an original equipment
manufacturer (OEM) telematics device installed during manufacture of the
vehicle
102, or an aftermarket telematics device connected, for example, through an On-
Board Diagnostics (OBD) port of, or wireless communications connection to, the
vehicle 102. Each of the one or more telematics devices 114 can be battery-
powered,
solar-powered, connected to the electrical system of the vehicle 102, or
combinations
of them. In some cases, the telematics devices 114 are capable of being
mounted in or
on the vehicle 102.
In some cases, the telematics devices 114 are capable of being untethered from
the vehicle 102 such that they are moveable within or outside of the vehicle.
In some
cases, the telematics devices 114 include a tag device placed or affixed in
the vehicle
102, such as tags of the kind described in U.S. Patent Application 14/529,812,
entitled
"System and Method for Obtaining Vehicle Telematics Data," filed October 31,
2014
and U.S. Patent Application 16/407,502, entitled "Safety for Vehicle Users,"
filed
May 9, 2019, the entire contents of which are incorporated by reference. In
some
cases, the telematics devices 114 include a mobile device, such as a
smartphone, a
wearable device, a tablet computer, a laptop computer, or another portable
computing
device, and may not necessarily be a telematics device dedicated to a
particular
vehicle 102. In some cases, the telematics devices 114 include a combination
of one
or more tag devices and one or more mobile devices. In some cases, the on-
board
computer 108 is capable of performing some or all of the functions of the
telematics
device 114.
In some implementations, the on-board computer 108 is capable of receiving
and processing the sensor data to operate the vehicle 102 or to assist the
driver 106
with operating the vehicle. In some implementations, the on-board computer 108
is
capable of receiving the sensor data from the sensors 112, the telematics
devices 114,
13
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
or both, and can store the sensor data in a memory 122. For example, in some
cases,
the on-board computer 108 interfaces with some or all of the sensors 112 using
a
wired or wireless communications interface 124 to receive the sensor data. In
some
cases, the telematics devices 114 are capable of interfacing with some or all
of the
sensors 112 using a wired or wireless communications interface 120 to receive
the
sensor data, and the on-board computer 108 and the telematics devices 114 can
establish a communications channel 128 between one another to exchange the
sensor
data and other information. The communications channel 128 can be a wired or
wireless communications channel, such as, among others, Bluetooth (TM), Wi-Fi
(TM), cellular, Radio Frequency Identification (RFID), Near-Field
Communication
(NFC), or combinations of them. In some cases, the on-board computer 108 or
the
telematics devices 114, or both, is capable of receiving sensor data or other
information from one or more components or devices in the surrounding
environment
104, such as infrastructure or an on-board computer 108 or a telematics device
114 in
another vehicle 102.
In some implementations, once received, one or more processors 126 of the
on-board computer 108 are capable of processing the sensor data in accordance
with
software instructions or applications stored in the memory 122 to control, for
example, the steering, throttle, or braking of the vehicle 102. To facilitate
such
control, the on-board computer 108 can be communicatively coupled with the
controls
or components of the vehicle 102 by various electrical or electromechanical
components. In some cases, such as those involving fully autonomous vehicles,
the
vehicle 102 is operable only by the on-board computer 108. In some cases, the
on-
board computer 108 supplement the controls or components operated by the
driver
106 of the vehicle 102, such as the steering wheel, the throttle pedal, or the
brake
pedal.
In various implementations, the technology 100 is capable of processing the
sensor data to evaluate the safety performance of the vehicle 102 or other
vehicles
within the environment 104. In some cases, the on-board computer 108, the
telematics
devices 114, or both, are capable of processing the sensor data to carry out
the safety
performance evaluation in accordance with the techniques described in this
specification. In some cases, the on-board computer 108, the telematics device
114, or
14
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
both, are capable of transmitting the sensor data to the server 130 for
processing. For
example, the on-board computer 108, the telematics devices 114, or both, can
establish a communications channel 132 with a network 134 to exchange the
sensor
data and other information with the server 130. The network 134 can be the
Internet, a
cellular network, a Wi-Fi network, a local area network, a wide area network,
a
satellite network, or any other suitable data transmission network, or
combinations of
them. Once received, the server 130 is capable of storing the sensor data and
other
information in a database 136. The database 136 can be implemented using one
or
more non-transitory computer-readable storage mediums including, but not
limited to,
hard disk drives, solid-state drives, optical storage drives, or combinations
of them.
In some implementations, the on-board computer 108, the telematics devices
114, or both, transmit the sensor data to the server 130 according to a
conditional data
upload protocol. For example, the on-board computer 108, telematics devices
114, or
both, may transmit only a portion of the sensor data to the server 130, while
.. maintaining other portions of the data in, for example, memory 118, 122.
The
transmitted portion of the sensor data can be data that is considered most
likely to be
useful in evaluating the safety performance of the vehicle 114. Later, if the
server 130
determines that the stored portion of data may be of interest, the on-board
computer
108, telematics device 114, or both, can transmit at least a portion of the
stored sensor
data to the server 130. The telematics devices 114 and the on-board computer
108 can
follow a similar conditional data upload protocol when transmitting data
between
each other. Transmitting data according to the conditional data upload
protocol can
alleviate concerns related to bandwidth constraints and computational power
constraints.
In some implementations, the sensor data is processed in accordance with a
tiered data processing protocol. For example, the sensor data can be processed
by the
telematics devices 114 according to a first tier data processing algorithm
while the
server 130 processes the data according to a second tier data processing
algorithm. In
such instances, the first tier data processing algorithm includes a simple,
less
computationally intensive algorithm (which can be less accurate) when compared
to
the second tier data processing algorithm. In some implementations, the tiered
data
processing protocol can include three tiers of data processing algorithms,
where one
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
telematics device 114 processes the data using a first tier algorithm, a
second
telematics device 114 processes data using a second tier algorithm, and the
server 130
processes data using a third tier algorithm. Each tier can include a more
computationally intensive and accurate algorithm when compared to a previous
tier.
Processing data in accordance with a tiered data processing protocol can
alleviate
concerns related to computational power constraints, bandwidth constraints,
and
computational speed constraints.
FIG. 2 is a block diagram depicting an example server 130 for evaluating
vehicle safety performance. The shown server 130 includes hardware and
software
components, such as one or more processors 200, a memory 202, and a
communication interface 204, which are interconnected by a data bus 206. The
memory 202 can be any non-transitory computer-readable storage medium and can
store computer-readable instructions executable by the processors 200. In
particular,
the memory 202 can store executable instructions associated with a risk
identification
module 208, a safety event identification module 210, a risk correlation
module 212, a
risk scoring module 214, a safety performance evaluation module 216, and other
modules, to enable the server 130 or other components and devices to carry out
the
techniques described in this specification. We use the term "module" broadly
to
include, for example, any code, program, firmware, software object, or other
software
device or arrangement that can be executed by one or more processors to
perform one
or more activities, functions, or facilities.
The risk identification module 208 is capable of processing the sensor data
produced at the vehicle 102 to identify the risk sources 110 encountered by
the
vehicle or other vehicles in the surrounding environment 104. In general, the
risk
identification module 208 can be configured to identify environmental risk
sources,
driver-specific risk sources, and vehicle-specific risk sources, although
additional,
fewer, or alternative risk sources may be identified in some implementations.
For
example, the risk identification module 208 can identify one or more risk
sources 110
due to features or conditions of the environment 104, such as the weather, the
time of
day, the level of lighting, the angle of lighting, the road type, the road
condition, the
area of the road in which the vehicle is operating (for example, an
intersection, a
merge lane, a shoulder, and so forth), the level of vehicle traffic, the level
of
16
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
pedestrian traffic, the speed of vehicles in the environment, the geography,
the
presence (or absence) of road markings, the presence (or absence) of
construction, the
proximity of the vehicle to other vehicles or objects, or the speed,
acceleration, or
behavior of other vehicles, or combinations of them, among others. In some
cases, the
risk identification module 208 is capable of identifying one or more risk
sources 110
due to an action (or inaction) by the driver 106 of the vehicle, such as
distracted
driving, driving while incapacitated, speeding, hard acceleration, hard
braking, hard
cornering, drifting, or swerving, or combinations of them, among others. In
some
cases, the risk identification module 208 is capable of identifying one or
more risk
sources due to damage to or malfunctioning of the vehicle 102.
To identify one or more risk sources 110 at a vehicle 102, the risk
identification module 208 is capable of processing the sensor data produced at
the
vehicle to detect, for example, the presence of a risk source at the vehicle
or the
exposure level of a risk source at the vehicle, or both, among others. One or
more risk
sources 110 may either be present or absent at the vehicle 102, and the risk
identification module 208 can process the sensor data to determine whether the
risk
source, and thus the risk, is present (or absent). For example, the risk
identification
module 208 can process image data from an image sensor at the vehicle 102,
such as a
road-facing camera, to detect the presence (or absence) of road construction
at the
vehicle 102. In some cases, one or more risk sources 110 may pose risk that
exists
along a continuum, and the risk identification module 208 is capable of
processing the
sensor data to determine the exposure level of the risk source at the vehicle
102. For
example, the level of lighting at the vehicle 102 may exist along a continuum,
with a
lower level of lighting generally corresponding to a higher risk of a safety
event due
to the reduced visibility. Accordingly, the risk identification module 208 can
process
the sensor data to identify the exposure level of the risk source 110 (e.g.,
the level of
lighting). In some cases, one or more risk sources 110 may include both
presence and
exposure level components. For example, the weather at the vehicle 102 can
have a
presence component, such as the presence of precipitation, and an exposure
level
component, such as the rate of precipitation. Accordingly, the risk
identification
module 208 can process the sensor data to identify both the presence of the
risk and
the exposure level of the risk source.
17
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
In some cases, the risk identification module 208 is capable of processing
other information in combination with the sensor data to identify one or more
risk
sources 110. In such instances, the risk identification module 208 is capable
of
receiving information, such as threshold information, from one or more
components
or devices of the technology 100 for use in determining whether the sensor
data
produced at the vehicle 102 indicates a source of risk. For example, to
determine
whether the speed of the vehicle 102 presents a source of risk, it may be
useful to
compare the speed data produced by one or more sensors 112 at the vehicle 102
with
the speed limit for the road on which the vehicle is traveling. Accordingly,
the risk
identification module 208 can receive the applicable speed limit data (e.g.,
based on
position data received from sensors at the vehicle) from, for example, the
database
136 or through a request to an application programming interface (API) or
other
service provided on a computing device 138, among others. Once received, the
risk
identification module 208 can compare the speed limit data with the speed data
produced at the vehicle to determine whether the speed of the vehicle 102 is a
source
of risk. In some cases, the risk identification module 208 is capable of
receiving
information from one or more components or devices of the technology 100 to
determine the presence or exposure level of a risk source 110 at the vehicle
102. For
example, the risk identification module 208 can use position data produced by
one or
more sensors at the vehicle 102 to request weather information from, for
example, a
weather service provided on a computing device 138 to identify the presence or
level,
or both, of a weather risk.
In some implementations, the risk identification module 208 is capable of
processing sensor data received from the vehicle 102 to identify the risk
sources 110
encountered by one or more other vehicles in the environment 104. For example,
the
risk identification module 208 may associate one or more risk sources 110
identified
for the vehicle 102 with other vehicles in the surrounding environment 104.
For
instance, if it is determined that there is a certain level of risk due to low
lighting at
the vehicle 102, then the risk identification module 208 may impute the same
risk to
other vehicles in the same environment 104 as the vehicle 102. In some cases,
the risk
identification module 208 is capable of using the sensor data produced at the
vehicle
102 to determine one or more risk sources 110 experienced by other vehicles in
the
18
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
environment 104. For example, the risk identification module 208 can process
sensor
data produced at the vehicle 102, such as radar, LIDAR, sonar, or image data,
among
others, to determine the distance, position, speed, or acceleration, or
combinations of
them, among others, of another vehicle in the environment 104. The risk
identification
module 208 may then use this data to identify one or more risk sources 110 for
the
vehicle 102 or the other vehicle in the environment 104, or both. These
identified risk
sources, and the risks they pose, can be incorporated into the risk score for
the vehicle
102 or used to determine a risk score for the other vehicle, or both, as
discussed
below. In this way, the technology 100 can identify risk sources, determine a
risk
score, and evaluate the safety performance of another vehicle in the
environment even
if the vehicle does not include sensors 112 or is not a participant in the
technology.
Further, by collecting data for both the vehicle 102 and other vehicles in the
surrounding environment 104 over the same time period, the technology 100 can
compare the risk score or safety performance, or both, for the vehicle and
other
vehicles. In doing so, the technology 100 can evaluate the relative safety
performance
of the vehicle 102 and other vehicles in the environment 104 to make
statistically
meaningful observations about overall safety performance of both the vehicle
102 and
the other vehicles in the environment 104.
The safety event identification module 210 is capable of processing sensor
data produced at a vehicle 102 to determine whether the vehicle or other
vehicles in
the surrounding environment 104 have experienced a safety event, such as a
vehicle
crash, a near-crash, or another unsafe condition. For example, the safety
event
identification module 210 may process sensor data produced at the vehicle 102,
such
as image data, audio data, speed data, or acceleration data, or combinations
of them,
among others, to identify one or more impacts that are characteristic of a
crash by the
vehicle or other vehicles. In some cases, the safety event identification
module 210 is
capable of processing sensor data to identify a near-crash of the vehicle 102
or other
vehicles, such as a situation that would have resulted in a crash but for an
evasive
maneuver by a vehicle or the driver 106 of the vehicle, or another unsafe
condition at
the vehicle or other vehicles, such as a vehicle driving off the road. In some
cases, the
safety event identification module 210 is capable of processing sensor data
produced
at the vehicle 102, such as radar, LIDAR, sonar, or image data, among others,
to
19
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
identify a safety event experienced by other vehicles in the environment 104.
In some
cases, the safety event identification module 210 is capable of receiving
information
regarding safety events from one or more other components or devices of the
technology 100, such as a crash record or crash report service provided on one
or
more computing devices 138. Techniques for identifying a vehicle crash, near-
crash,
or another unsafe condition at the vehicle are described in U.S. Patent
Application
16/035,861, entitled "Vehicle Telematics of Vehicle Crashes," filed July 16,
2018, the
entire contents of which is incorporated by reference.
After identification, the risk sources 110 and safety events experienced by
the
vehicle 102 or other vehicles in the surrounding environment 104 can be
reported to
one or more other modules, such as the risk correlation module 212, the risk
scoring
module 214, and the safety performance evaluation module 216. For example, the
risk
identification module 208 can report the presence or exposure level, or both,
of each
risk source 110 to one or more other modules, and the safety event
identification
module 210 can report the occurrence or type, or both, of each safety event to
one or
more other modules. In some cases, the risk identification module 208 or the
safety
event identification module 210, or both, are capable of providing a timestamp
or
another marker in the sensor data of the identified risk sources 110 or safety
events to
the other modules in addition to, or in lieu of, reporting the identifications
directly.
To accurately determine the risk of a safety event experienced by a vehicle
102, it can be useful to understand the risk of a safety event posed by each
risk source
110 or set of risk sources. Accordingly, the risk correlation module 212 may
process
the sensor data produced at the vehicle 102, the identified risk sources 110
received
from the risk identification module 208, and the identified safety events
received from
the safety event identification module 210 to establish a correlation between
each risk
source 110 or set of risk sources and the occurrence of a safety event. This
correlation
may be used to determine the risk of a safety event posed by each risk source
110 or
set of risk sources or combined with other information, such as presence and
exposure
level information, to determine the risk posed by each risk source or set of
risk
sources. The risk correlation module 212 can aggregate the sensor data,
identified
risks, and identified safety events received from one or more vehicles 102 or
types of
vehicles over many miles to increase the sample size and establish the
correlation
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
with greater accuracy. Further, the risk correlation module 212 can
continually update
the correlation as new sensor data is received from one or more of the
vehicles 102.
The risk correlation module 212 is capable of using one or more algorithms to
establish the correlation between each risk source or set of risk sources and
a safety
event. For example, the risk correlation module 212 can determine how often a
risk
source or set of risk sources is present during the occurrence of a safety
event to
establish the correlation. In some cases, the risk correlation module 212 is
capable of
determining which risk source or set of risk sources is the actual or
proximate cause
of the safety event to establish the correlation. In some cases, the risk
correlation
module 212 is capable of accounting for the exposure level of a risk source or
set of
risk sources during the occurrence of a safety event to establish the
correlation. In
some cases, the risk correlation module 212 is capable of using machine
learning
techniques to establish the correlation. In some cases, the risk correlation
module 212
is capable of determining a different correlation for a particular risk source
when it
appears with one or more other risk sources than when it appears alone to
account for
the accumulative effect of certain risks. To increase the strength of the
correlation
under various circumstances, the risk correlation module 212 may establish
multiple
correlation matrices specific to, for example, a vehicle type, a particular
safety event,
or a particular environment 104, among others.
Once a correlation is established, the risk scoring module 214 is capable of
using the correlation and the sensor data to derive a risk score for the
vehicle 102 or
other vehicles in the environment 104. The risk score for each vehicle can
represent
the risk of a safety event experienced by the vehicle during a particular
period of
operation, such as a particular distance traveled or time operated, based on
the actual
risk sources encountered by the vehicle and the risk of a safety event posed
by each
risk. In this way, the risk score can serve as an accurate measure of the risk
of a safety
event experienced by the vehicle during the period of operation.
To determine the risk score for the vehicle 102 or other vehicles in the
environment 104, the risk scoring module 214 is capable of receiving from, for
example, the risk identification module 208 an indication of the risk sources
110
encountered by the vehicle during a particular period of operation. The period
of
operation can include a particular distance traveled, such as a mile or any
other
21
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
distance, or a particular time operated, such as five minutes or any other
time, among
others. The risk scoring module 214 can use the correlation information
received from
the risk correlation module 212 to determine the risk posed by each risk
source. In
some cases, the risk scoring module 214 is capable of applying a weight to
some or all
of the identified risk sources 110 based on the exposure level of the risk
source or the
presence (or absence) of the risk source or other risk sources, or both, among
others.
The risk scoring module 214 can determine the risk score for the particular
period of
operation by, for example, calculating a sum of the risk or the weighted risk
posed by
each risk source. In some cases, the risk scoring module 214 is capable of
normalizing
the risk score to a standard risk for a particular period of operation. For
example, as
indicated previously, the risk scoring module 214 can determine an average or
median
level of risk experienced by a vehicle over a unit distance traveled, such as
a mile
under standard risk conditions, and can normalize each risk score so that high-
risk
travel receives an above-normal risk score and low-risk travel receives a
below-
normal risk score.
In some cases, the risk scoring module 214 (or another module, such as the
risk scoring module 212) is capable of filtering, discounting, or ignoring
certain risk
sources or types of risk sources when determining the risk score. For example,
when
comparing the safety performance of autonomous vehicles and conventional
vehicles,
it may be useful to filter out driver-specific risk sources or vehicle-
specific risk
sources, or both, so that the safety performance of each vehicle is evaluated
based on
how the vehicle 102 or the driver 106 of the vehicle responds to environmental
risk
sources. In this way, the technology 100 can compare, for example, the safety
performance of autonomous vehicles and conventional vehicles without biases
due to
risks created by driver-specific behavior or vehicle-specific problems.
The safety performance evaluation module 216 (which is sometimes referred
to as the "safety performance module") is capable of using the risk score and
the
identified safety events to evaluate the safety performance of a vehicle or
vehicle
type. For example, the safety performance module 216 can receive a risk score
for a
particular period of operation, such as a mile, from the risk scoring module
214. The
safety performance module 216 can use information received from the safety
event
identification module 210 to determine the number of safety events experienced
by
22
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
the vehicle over the same period for which the risk score was calculated. The
safety
performance module 216 can then calculate the number of safety events
experienced
by the vehicle relative to the risk score for the vehicle to determine the
vehicle's
safety performance for the particular period of operation (e.g., the number of
safety
events divided by the risk score). The safety performance of the vehicle may
be
represented as a ratio, a percentage, a score, a ranking, or any other metric,
and may
be communicated to one or more components or devices of the technology 100,
such
as the on-board computer 108, the one or more telematics devices 114, the
database
136, or a computing device 138, or combinations of them, as discussed below.
In some cases, the safety performance evaluation module 216 is capable of
combining multiple risk scores and safety events to evaluate the safety
performance of
a vehicle or vehicle type over multiple periods of operation. In some cases,
the safety
performance module 216 is capable of combining the risk scores from, for
example,
each mile of a trip by a vehicle to determine a risk score for the entire
trip. We use the
term "trip" broadly to include, for example, any instance of travel from an
origin
place to a destination place. In some cases, the safety performance module 216
may
combine the risk scores for each mile traveled by a vehicle or vehicle type to
determine a risk score for the cumulative distance traveled by the vehicle or
vehicle
type. The safety performance module 216 is capable of determining the combined
number of safety events experienced by the vehicle or vehicle type over the
same
operation period for which the combined risk score was calculated. The safety
performance module 216 can then calculate the combined number of safety events
experienced by the vehicle or vehicle type relative to the combined risk score
for the
vehicle or vehicle type to determine the safety performance for the vehicle or
vehicle
.. type. By determining the safety performance of a vehicle or vehicle type
over multiple
operation periods, the safety performance module 216 can attain more accurate
and
statistically significant performance metrics that can be compared among
vehicles or
types of vehicles.
In some cases, the safety performance evaluation module 216 is capable of
applying the risk score to determine a risk-normalized distance traveled by a
vehicle
or vehicle type. For example, in cases where the risk score is calculated in
mile
increments, the safety performance module 216 can apply the risk score for a
23
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
particular mile to determine a risk-normalized mile for the vehicle or vehicle
type. In
this way, the distance traveled by the vehicle or vehicle type can be adjusted
for safety
performance evaluation purposes to provide greater weight to high-risk travel
and less
weight to low-risk travel. In some instances, determining a risk-normalized
mile
involves comparing the distance traveled to a typical risk-normalized mile
(such as
the "yearly average" mile or the "Cambridge" mile discussed previously). In
some
cases, the safety performance module 216 may combine the risk-normalized
distances
for a particular vehicle or vehicle type to determine a total risk-normalized
distance
traveled by the vehicle or vehicle type. The safety performance module 216 can
calculate the number of safety events experienced over the actual distance
traveled by
the vehicle or vehicle type relative to the risk-normalized distance to
determine the
safety performance of the vehicle or vehicle type. By applying the risk score
to
determine a risk-normalized distance traveled, the safety performance metrics
described in this specification can be compared with conventional metrics for
vehicle
safety performance, such as the number of safety events per vehicle-mile-
traveled
(VMT).
After determining the safety performance of a vehicle or vehicle type, the
safety performance evaluation module 216 is capable of analyzing the safety
performance and comparing it with the safety performance of other vehicles or
types
of vehicles. For example, the safety performance module 216 can compare the
safety
performance of different vehicles or types of vehicles to determine a ranking
based on
the relative safety performance of each vehicle or vehicle type. In some
cases, the
safety performance module 216 is capable of comparing the safety performance
of
different vehicles or types of vehicles to make relative claims about the
operators of
the vehicles or vehicle types. For example, the safety performance module 216
can
compare the safety performance of autonomous vehicles with that of
conventional
vehicles to determine the relative safety performance of the human and
computer-
implemented operators of such vehicles. In some cases, the safety performance
module 216 is capable of comparing the safety performance of a particular
vehicle
102 or driver 106 with one or more previous safety performance evaluations to
determine the safety performance of the vehicle or the driver over time.
24
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
In some cases, the safety performance module 216 is capable of comparing the
safety performance of two or more vehicles that are traveling in the same
environment
104 at the same time. For example, the technology 100 can use sensor data
acquired at
a vehicle 102 to determine the safety performance of both the vehicle 102 and
other
vehicles in the surrounding environment 104, as discussed above.
Alternatively,
additionally, the technology 100 can use sensor data acquired by separate
sensors 112
at each vehicle 102 in the environment 104 to determine the safety performance
of
each vehicle. Because each vehicle is likely facing the same or substantially
similar
risks due to their proximity in the environment 104, strong comparisons of the
relative
safety performance of each vehicle or vehicle operator can be made.
In some implementations, the safety performance evaluation module 216 can
provide the safety performance evaluation to one or more components or devices
of
the technology 100. For example, in some cases, the safety performance module
216
is capable of causing the server 130 to provide the safety performance
evaluation to
.. the database 136 for storage, to the on-board computer 108 or the
telematics devices
114 over the network 134, or to one or more computing devices 138 over a
network
140, which can be the Internet. Each of the computing devices 138 can be
associated
with a user or owner of the vehicle 102, a user or owner of the telematics
devices 114,
a vehicle safety organization, an vehicle producer, an insurance company, a
ridesharing company, an emergency service, a call center, an analyst, a
researcher, a
host of the technology, or combinations of these participants, among others.
The
server 130 can provide the safety performance evaluation to the components or
devices through a webpage, an application, an API, or a notification, or
combinations
of them, among others. Once received, each of the components or devices may
present the safety performance evaluation to a user of the component or device
using,
for example, a graphical user interface presented on a display of the
component or
device.
In general, the safety performance evaluation module 216 can provide any
aspect of the safety performance evaluation to the other components or
devices,
including one or more of the identified risk sources 110, the identified
safety events,
the risk score, or the determined safety performance, or combinations of them,
among
others. In some cases, the safety performance module 216 is capable of
providing one
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
or more comparisons of the safety performance for the vehicle with the safety
performance of the same or different vehicles. In some cases, the safety
performance
module 216 is capable of providing recommendations or suggestions to increase
the
safety performance of the vehicle, such as suggestions to the vehicle 102 or
the driver
106, or both, to operate the vehicle in a safer manner or recommendations to
the
vehicle or driver, or both, to follow a safer route in future trips. In some
cases, the
safety performance module 216 is capable of indicating a time or distance,
such as a
number of risk-normalized miles, that a particular vehicle or vehicle type
needs to
travel to attain a certain level of safety performance. For example, the
safety
performance module 216 can indicate a number of risk-normalized miles that an
autonomous vehicle may need to travel to achieve a comparable or higher safety
performance than conventional vehicles and human drivers.
In some cases, the safety performance evaluation can be used by government
agencies in certifying the safety of, for example, an autonomous vehicle. In
some
.. cases, the safety performance evaluations can be used to improve the safety
of
algorithms used for navigating an autonomous vehicle. For example, algorithms
used
to control an autonomous vehicle can be updated in a manner to improve the
safety
performance evaluation. These updates can be used to improve the evaluation of
driving safety or to improve the actual driving safety, or a combination of
the two,
among other things. In some cases, the safety performance evaluation can be
used to
determine one or more routes that should be taken during a vehicle's commute
to
reduce environmental risks.
Various modifications to the safety performance evaluation technology 100 are
possible. For example, in some cases, the technology 100 can average multiple
risk
scores from one or more vehicles or types of vehicles over the same distance
traveled
to determine an average risk score for the particular distance. In this way,
the
technology 100 can identify high-risk routes, which may allow the technology
100 to
suggest safer alternatives to the vehicle 102 or the driver 106, or both,
among others.
In some cases, the technology 100 can apply the risk score to normalize other
risk
exposure metrics, such as time driven, and is not limited to normalizing
distance
traveled.
26
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
Although the server 130 is described as processing the sensor data associated
with the vehicle 102 to evaluate the safety performance of the vehicle or
other
vehicles within the environment 104, other components and devices of the
technology
100, such as the on-board computer 108, the telematics devices 114, or the
computing
device 138, or combinations of them, may process the sensor data in addition
to, in
combination with, or instead of the server 130 to carry out the techniques
described
here. Similarly, although specific modules, including the risk identification
module
208, the safety event identification module 210, the risk correlation module
212, the
risk scoring module 214, and the safety performance evaluation module 216 are
.. described as carrying out certain aspects of the techniques described here,
some or all
of the techniques may be carried out by additional, fewer, or alternative
modules in
some implementations. Further, although only one server 130, one database 136,
and
one computing device 138 are illustrated in figure 1, the technology 100 may
include
any number of computing devices and data storage devices located in a single
place or
.. distributed and communicatively connected using any number of networks.
Figure 3 illustrates an example process 300 for evaluating the safety
performance of a vehicle. At least a portion of the process 300 can be
implemented
using one or more processors, such as the processors 200 operating on the
server 130.
Operations of the process 300 include receiving sensor data produced by one or
sensors at a vehicle (302). The sensors may be installed in the vehicle or
included in a
telematics device brought into the vehicle, or both.
The sensor data is processed to identify one or more risk sources experienced
by the vehicle during a period of operation (304). The risk sources may put
the
vehicle at risk of a safety event. In some cases, the one or more risk sources
may
.. include environmental risk sources due to conditions or features of the
environment in
which the vehicle travels over the distance. In some cases, the environmental
risk
sources may include other vehicles in the environment. In some cases, driver-
specific
risk sources due to conduct of a driver of the vehicle or vehicle specific-
risk sources
due to damage to or malfunctioning of the vehicle, or both, may be filtered
from the
.. one or more risk sources. In some cases, the period of operation may be
associated
with a distance traveled by the vehicle or a time of operation of the vehicle.
27
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
The sensor data is processed to identify one or more safety events experienced
by the vehicle during the period of operation (306). In some cases, the one or
more
safety events may include a crash of the vehicle, a near-crash of the vehicle,
or
another unsafe condition at the vehicle, or combinations of them.
A risk of a safety event experienced by the vehicle during the period of
operation is determined based on the one or more risk sources (308). In some
cases,
the risk can be measured by a risk score. In some cases, determining the risk
may
include determining a risk of a safety event posed by each of the risk
sources, the
exposure-level of the risk source, or the presence of the risk source, or
combinations
of them, among others. The risk of a safety event posed by each risk source
can be
summed to determine the risk for the vehicle during the period of operation.
In some
cases, the risk may be normalized based on a standard risk of a safety event
experienced by the vehicle or other vehicles during the period of operation.
In some
cases, the risk score may be applied to the period of operation to produce a
risk-
normalized period of operation, such as a risk-normalized distance traveled by
the
vehicle. In some cases, the risk of a safety event experienced by the vehicle
may be
combined with other risks of a safety event experienced by the vehicle or
other
vehicles having a same vehicle type to produce a combine risk of a safety
event for
the vehicle or the vehicle type.
The safety performance of the vehicle during the period of operation is
determined based on the one or more safety events and the risk (310). In some
cases,
determining the safety performance of the vehicle may include determining the
sum
of the one or more safety events relative to the risk. In some cases, the risk
may be
applied to the period of operation to produce a risk-normalized period of
operation,
such as a risk-normalize distance traveled by the vehicle, and the safety
performance
of the vehicle during the period of operation can be determined based on the
sum of
the one or more safety events relative to the risk-normalized period of
operation. In
some cases, the sum of the one or more safety events is a weighted sum. In
some
cases, each of the safety events of the one or more safety events are weighted
according to the severity of the safety event. For example, a safety event
that results
in a totaled vehicle can be weighted more heavily than a safety event that
resulted in
minor paint damage to the vehicle. In some cases, the risk of a safety event
28
CA 03147603 2022-01-14
WO 2021/016035
PCT/US2020/042320
experienced by the vehicle may be combined with other risks of a safety event
experienced by the vehicle or other vehicles having a same vehicle type to
produce a
combined risk of a safety event for the vehicle or the vehicle type, and the
safety
performance of the vehicle or vehicle type can be determined based in part on
the
combined risk of a safety event.
In some implementations, the sensor data may be processed to identify one or
more risk sources and safety events experienced by a second vehicle in a same
environment as the vehicle during the period of operation. A risk of a safety
event
experienced by the second vehicle during the period of operation may be
determined
based on the one or more risk sources encountered by the second vehicle. The
safety
performance of the second vehicle may be determined during the period of
operation
based on the sum of the one or more safety events experienced by the second
vehicle
relative to the risk for the second vehicle. In some cases, a relative safety
performance
of the vehicle during the period of operation can be determined by comparing
the
safety performance of the vehicle with the safety performance of the second
vehicle.
Other implementations are also within the scope of the following claims.
29
